Two former Palantir employees reconnected by phone. The greeting wasn't "hello" — it was "Are you tracking Palantir's descent into fascism?" The people building the surveillance infrastructure are the ones sounding the alarm.
Canada and Germany are backing a Cohere-Aleph Alpha merger to build "sovereign AI." The combined entity is valued at $20 billion. Its annual revenue: roughly $240 million.
Police didn't have a suspect — just a map and a warrant served on Google. The Supreme Court will decide if that's enough to search the location history of every person who happened to be nearby.
Economy World launched yesterday on Steam's New Releases chart. Its first two players both report it demanded their Steam login credentials in a separate window. At time of writing, it is still available for download.
Anthropic traced 16 million fraudulent queries to three Chinese AI labs through 24,000 fake accounts. Now the State Department has told every embassy on Earth to sound the alarm.
A sophisticated backdoor survived seven months of patching on an unnamed federal agency's Cisco firewall. CISA and the UK's NCSC published the technical details — but not the victim's identity.
OpenAI flagged and banned Jesse Van Rootselaar's ChatGPT account in June 2025. She killed eight people in February 2026. The company never contacted law enforcement.
A malware framework from 2005 silently corrupted physics and engineering simulations across entire facilities — predating Stuxnet by five years and raising questions about whether decades of safety-critical calculations can be trusted.
DeepSeek V4 processes a million words at a time, runs on Huawei's Chinese-made chips, and costs a fraction of what OpenAI charges. Washington noticed.
OpenAI's fourth model in five months brings measurable token efficiency gains and the company's clearest super app blueprint yet. The NVIDIA dependency underneath got zero airtime.
Scammers posed as Iranian authorities and demanded cryptocurrency for safe passage through the Strait of Hormuz. At least one vessel that fell for it was fired upon by the IRGC.
Half a million UK volunteers gave their health data to science. Someone with legitimate access then listed it on a Chinese shopping website. No hack required.
Intel's CEO warned he'd exit contract chipmaking without a major 14A customer. Elon Musk just handed him one — though the Austin Terafab complex it's tied to remains long on vision and short on specifics.
Tesla sold four million cars whose hardware it promised could drive itself. The hardware can't. Now Musk wants to build factories across the country to swap the computers out.
The National Agency for Secure Titles — the word is right there in the name — may have lost data covering a third of France's population. Criminals are already shopping the haul on underground forums.
TSMC can now build cutting-edge chips without ASML's $380 million high-NA machines — and Google just replaced Intel's x86 with its own Arm silicon across its entire AI accelerator line. The foundational hardware of the AI era is being redesigned by an ever-smaller circle of companies.
The law firm that advises OpenAI on responsible AI deployment filed fabricated citations in federal bankruptcy court. Its safeguards didn't catch them. Opposing counsel did.
Florida's attorney general says if ChatGPT were a person, he'd charge it with murder. He's now investigating whether OpenAI bears criminal culpability all the same — a first in US legal history.
An unauthorized group accessed Anthropic's most dangerous cybersecurity AI on launch day through a vendor loophole. The federal agency charged with protecting America's critical infrastructure still doesn't have it.
CATL's latest LFP battery charges from 10% to 98% in just over six minutes — roughly three minutes faster than BYD's best. No Western battery maker has publicly demonstrated anything close.
lop News
Jonny V. Nuovo